White Papers

Known issues and resolution
17 Dell EMC PowerEdge Servers with NVIDIA GPUs and VMware vSphere | Technical white paper
7. On the PowerEdge R740 with NVIDIA Tesla T4 and installed with ESXi 6.7 and attempt to list the NVIDIA
GPU in Ispci command fails.
- Resolution:
This behavior may be due to multiple reasons. Check the following:
• Ensure that both the PSUs are plugged-in and working.
• Ensure that the correct wattage of PSU is used for the GPU configuration.
• Ensure that the GPU power cables are connected.
• Ensure that the GPU is not configured as a pass-through device in the ESXi host.
8. On the PowerEdge R730 server with Tesla M60, the VM configured with Tesla M60 for vDGA displays a
blank screen or fails to power-on.
- Resolution:
The Tesla M60 can work in both graphics mode and computation mode. Ensure that, the NVIDIA M60
GPU is configured in graphics mode and not in computation mode. For toggling the modes, use the tool
gpumodeswitch. For more information, see the document GPUMODESWITCH User Guide at NVIDIA
support site.
9. On the PowerEdge R740 with Tesla T4 and ESXi 6.5, VM configured with multi vGPU fails to power on or
fails to initialize vGPU after boot.
- Resolution:
Assigning multiple vGPUs is not supported in ESXi 6.5. Either update the host to ESXi 6.7 U3 or ensure
that the VM has only one vGPU associated to it.
10. On the PowerEdge R740 server with Tesla T4 and ESXi 6.5 U1, the VM with vGPU associated to it fails
to boot.
- Resolution:
The VM associated with vGPU power-on fails with the error message “The amount of graphics resource
available in the parent resource pool is insufficient for the operation.” This behavior is seen if the VM
Graphics Type is set to Shared. Change the Graphics Type to the Shared Direct option. Note that the
default option is set to Shared.
11. On a PowerEdge R730 with M60 GPU and ESXi 6.5 U3, the VM failed to power-on. The VM log file
contains the following entry:
2019-10-07T06:57:51.499Z| vmx| I120: PCIPassthru: total number of pages needed (2097186) exceeds
limit (917504), failing
2019-10-07T06:57:51.499Z| vmx| I120: Module DevicePowerOn power on failed
- Resolution:
In order to resolve this issue, either reduce the memory assigned to the VM and power on, or perform the
followings:
• Ensure that the BIOS configuration setting Memory Mapped I/O above 4GB is set to Enable.
• Add the below command in the VMX file of VM:
o pciPassthru.use64bitMMIO=”TRUE”
o pciPassthru.64bitMMIOSizeGB = “64”